CN114554175A - Point cloud lossless compression method based on classification rearrangement - Google Patents

Abstract

The invention relates to a point cloud lossless compression method based on classification rearrangement. First, a threshold value is determined: dividing pixels in an image into a plurality of categories so that the gray values of the pixels in the same category are close to each other; according to the determined threshold, C1, C2 and a pixel mask map Mp are obtained through classification; mp is a binary pixel mask map with size SxS; mp records the category information of all pixels, but the bit number occupied by the Mp is large, and further blocks the pixels are combined to obtain a block mask map Mb; traversing the block mask graph and the pixels according to the sequence of a third-order Hilbert curve, namely arranging all the pixels of the whole image into a one-dimensional sequence D, wherein the D sequence is divided into two subsequences D1 and D2 which respectively correspond to high and low gray values; the data streams D1 and D2 and the block mask diagram Mb are coded by JPEG-LS, and then two coded bit streams are formed and finally output. The invention can meet the dual requirements of data lossless and high compression ratio.

Description

Point cloud lossless compression method based on classification rearrangement

Technical Field

The invention relates to an image compression coding technology, in particular to a point cloud lossless compression method based on classification rearrangement.

Background

With the development of 3D sensing and capturing technologies, 3D lidar point cloud technologies have demonstrated their utility in digital perception of real world environments, such as virtual/augmented/mixed reality (AR/VR/MR), moving maps, historical cultural relic scanning, smart cities, robots, and autopilots. The point cloud data has a data volume, which is a great challenge for the storage and transmission of computer equipment information at present, and the effective compression thereof is one of the indispensable steps.

Image coding can be classified into lossy compression coding and lossless compression coding according to whether there is a loss of information in the coding process. Although lossy compression encoding can compress data to a great extent, due to information loss, a decoding end cannot restore and recover an original image. Lossy coding is undesirable when there is a need for data integrity. Lossless compression coding is a method in which there is no loss of information and the original image can be recovered without error at the decoding end. However, the current lossless compression coding generally has the problem that the compression rate is not high enough.

The point cloud data are rearranged according to the sequence of Hilbert curves, similar pixel values are gathered in a certain sense, spatial correlation is increased, the capacity of an image compression method is fully exerted, and a JPEG-LS data encoding method is utilized, so that lossless compression of an image can be realized, and the compression ratio can be ensured to be high enough.

Disclosure of Invention

The invention aims to provide a point cloud lossless compression method based on classification rearrangement, and aims to improve the compression efficiency of point cloud image coding and meet the application requirement of efficiently and accurately coding a point cloud image.

In order to achieve the purpose, the technical scheme of the invention is as follows: a point cloud lossless compression method based on classification rearrangement comprises the following encoding processes:

step S1, determining a threshold: dividing pixels in an image into a plurality of categories so that the gray values of the pixels in the same category are close to each other; firstly, adaptively determining a threshold value according to a histogram of an image to enable the quantity of two types of pixels to be close; let the histogram of the original image be: h ═ c (g) e [0, S²]And c (g) e Z, g 0, 1.., 65536 }; the threshold G should be chosen to be G-argmin_T{|∑_g＜TC(g)-∑_h≥TC (h) | }, wherein G, T and h are values of the histogram, and finding T to minimize the formula is the determined threshold value G;

step S2, obtaining C1, C2 and a pixel mask map Mp by gray classification according to the determined threshold: if the pixel value I (I, j) is larger than or equal to G, I (I, j) belongs to C1, and Mp (I, j) is 1; if the pixel value I (I, j) < G, I (I, j) ∈ C2 and Mp (I, j) > 0, C1 and C2 respectively represent the sets of pixels in the two categories, I (I, j) < G;

step S3, obtaining a block mask map Mb by merging mps: mp is a binary pixel mask map, size SxS; mp records the category information of all pixels, but the Mp itself has a large number of bits, and further combines the blocks, each block having a size MxM (M2M, M2, 3.) to obtain a block mask map Mb; when M is 1, Mp is equal to Mb; when the excessive pixel value of the area pixel value of 1 in the Mp corresponding to the pixel in Mb is 0, setting the pixel in Mb to be 1, otherwise setting the pixel to be 0;

step S4, inter-block rearrangement: traversing the block mask diagram Mb according to the sequence of the third-order Hilbert curves, and arranging all blocks into a one-dimensional sequence;

step S5, intra block rearrangement: each block also contains MxM pixels, so that the pixels are rearranged in the block by using Hilbert, all the blocks are performed in the same way, all the pixels of the whole image can be arranged into a one-dimensional sequence D, and the D sequence is divided into two subsequences D1 and D2 which respectively correspond to high and low gray values;

step S6, directly adopting JPEG-LS to encode the subsequences D1 and D2;

step S7, as each pixel of the block mask map Mb only contains a binary value, firstly combining 8 continuous pixels into a byte, and then carrying out JPEG-LS coding on the obtained byte stream;

the two encoded bitstreams are formed through steps S6 and S7 and finally output.

In an example of the present invention, the method further includes a decoding process of:

m1, recovering D1, D2 and Mb by using a JPEG-LS decoding algorithm;

step M2, traversing Mb according to the sequence of inter-block Hilbert curves, and taking M multiplied by M pixels from D1 or D2 according to the Mb value of the current point each time;

m3, filling pixels on the current block of the original image according to the sequence of Hilbert curves in the block; once the traversal is completed, the original image is obtained.

Compared with the prior art, the invention has the following beneficial effects: the invention rearranges the data of the two-dimensional point cloud distance image on the basis that the three-dimensional point cloud is converted into the two-dimensional distance image. The point cloud data are divided into high gray value and low gray value through threshold judgment, the pixel mask map is favorably reduced in occupied space by the inter-block rearrangement and the intra-block rearrangement, the point cloud data are rearranged according to the sequence of Hilbert curves, similar pixel values are gathered in a certain sense, the spatial correlation is increased, and the capacity of the image compression method is fully exerted, so that the double advantages of data loss and high compression ratio can be realized.

Drawings

FIG. 1 is a flow chart of the encoding of the present invention.

FIG. 2 is a decoding flow chart of the present invention.

Fig. 3 is a Hilbert graph.

Detailed Description

The technical scheme of the invention is specifically explained below with reference to the accompanying drawings.

As shown in fig. 1-3, the invention relates to a point cloud depth image lossless compression method based on classification rearrangement, which comprises the following steps:

1) the encoding process is shown in the encoding flow chart of fig. 1:

and step S1, determining a threshold value. Pixels in an image are divided into several categories such that the pixel grayscale values in the same category are close. Firstly, a threshold value is determined according to the histogram adaptation of the image, so that the number of the two types of pixels is close. Let the histogram of the original image be: h ═ c (g) e [0, S²]And c (g) ∈ Z, g ═ 0, 1, ·, 65536 }; . The threshold G should be chosen to be G-argmin_T{|∑_g＜TC(g)-∑_h≥TC (h) | }, where G, T, and h are values of the histogram, and finding a suitable T to minimize the equation, i.e., the threshold G.

Step S2, gray level classification is carried out to obtain C1, C2 and a pixel mask map Mp. If the pixel value I (I, j) is larger than or equal to G, I (I, j) belongs to C1, and Mp (I, j) is 1; if the pixel value I (I, j) < G, I (I, j) ∈ C2 and Mp (I, j) > 0, C1, C2 respectively represent the sets of pixels in the two categories of high grayscale and low grayscale;

step S3, a block mask map Mb is obtained by merging mps. Mp is a binary pixel mask map, size SxS. Mp records the category information of all pixels, but the Mp itself occupies a large number of bits, and we further merge the blocks, each block having a size MxM (M2M, M2, 3.) to obtain a block mask map Mb. When M is 1, Mp is equal to Mb. When the area pixel value in Mp corresponding to the pixel in Mb is more than 0 pixel value of 1, setting the pixel in Mb as 1, otherwise setting the pixel as 0;

and step S4, inter-block rearrangement. Traversing the block mask map Mb according to the sequence of the third-order Hilbert curves (shown in FIG. 3), and arranging all blocks into a one-dimensional sequence;

step S5, block rearrangement, each block also contains MxM pixels, so the block rearrangement is carried out by Hilbert, all the blocks are carried out in the same mode, all the pixels of the whole image can be arranged into a one-dimensional sequence D, the D sequence is divided into two subsequences D1 and D2, and the subsequences are respectively corresponding to high gray values and low gray values;

step S6, directly coding the data streams D1 and D2 by JPEG-LS;

step S7, as each pixel of the block mask map Mb only contains a binary value, firstly combining 8 continuous pixels into a byte, and then carrying out JPEG-LS coding on the obtained byte stream;

the two encoded bitstreams are formed through steps S6 and S7 and finally output.

2) The decoding process is shown in the decoding flow chart of fig. 2:

m1, recovering D1, D2 and Mb by using a JPEG-LS decoding algorithm;

step M2, traversing Mb according to the sequence of inter-block Hilbert curves, and taking M multiplied by M pixels from D1 or D2 according to the Mb value of the current point each time;

step M3, filling pixels onto the current block of the original image in the order of the Hilbert curves within the block. Once the traversal is complete, the original image is obtained.

Although the present invention has been described with reference to the preferred embodiments, it is not intended to limit the present invention, and those skilled in the art can make variations and modifications of the present invention without departing from the spirit and scope of the present invention by using the methods and technical contents disclosed above. The above description is only a preferred embodiment of the present invention, and all equivalent changes and modifications made in accordance with the claims of the present invention should be covered by the present invention.

Claims (2)

1. A point cloud lossless compression method based on classification rearrangement is characterized by comprising the following encoding processes:

step S1, determining a threshold: scribing pixels in an imageDividing the image into a plurality of categories to enable the gray values of pixels in the same category to be close; firstly, adaptively determining a threshold value according to a histogram of an image to enable the quantity of two types of pixels to be close; let the histogram of the original image be: h ═ c (g) e [0, S²]And c (g) e Z, g 0, 1.., 65536 }; the threshold G should be chosen to be G-argmin_T{|∑_g＜TC(g)-∑_h≥TC (h) | }, wherein G, T and h are values of the histogram, and finding T to minimize the formula is the determined threshold value G;

step S2, obtaining C1, C2 and a pixel mask map Mp by gray classification according to the determined threshold: if the pixel value I (I, j) ≧ G, I (I, j) belongs to C1 and Mp (I, j) equals 1; if the pixel value I (I, j) < G, I (I, j) ∈ C2 and Mp (I, j) > 0, C1 and C2 respectively represent the sets of pixels in the two categories, I (I, j) < G;

step S3, obtaining a block mask map Mb by merging mps: mp is a binary pixel mask map, size SxS; mp records the category information of all pixels, but the Mp itself occupies a large number of bits, and further merges the blocks, each block having a size MxM (M2M, M2, 3.) to obtain a block mask map Mb; when M is 1, Mp is equal to Mb; when the excessive pixel value of the area pixel value of 1 in Mp corresponding to the pixel in Mb is 0, setting the pixel in Mb to be 1, otherwise setting the pixel to be 0;

step S4, inter-block rearrangement: traversing the block mask diagram Mb according to the sequence of the third-order Hilbert curves, and arranging all blocks into a one-dimensional sequence;

step S5, intra block rearrangement: each block also contains MxM pixels, so that the pixels are rearranged in the block by using Hilbert, all the blocks are performed in the same way, all the pixels of the whole image can be arranged into a one-dimensional sequence D, and the D sequence is divided into two subsequences D1 and D2 which respectively correspond to high and low gray values;

step S6, directly adopting JPEG-LS to encode the subsequences D1 and D2;

step S7, as each pixel of the block mask map Mb only contains a binary value, firstly combining 8 continuous pixels into a byte, and then carrying out JPEG-LS coding on the obtained byte stream;

the two encoded bitstreams are formed through steps S6 and S7 and finally output.

2. The method of claim 1, further comprising a decoding process of:

step M1, recovering D1, D2 and Mb by using a JPEG-LS decoding algorithm;

step M2, traversing Mb according to the sequence of inter-block Hilbert curves, and taking M multiplied by M pixels from D1 or D2 according to the Mb value of the current point each time;

m3, filling pixels on the current block of the original image according to the sequence of Hilbert curves in the block; once the traversal is completed, the original image is obtained.

Images